The underlying science for the chemical and biological changes resulting from exposure to electron and photon beams is well understood. A significant world business which treats several billions of dollars of product annually, has ben created by the exploitation of radiation technology. In general, electron accelerators are used to process biologically inert materials to improve the physical characteristics of materials while intense radiation sources omitting higher penetration photons are used to sterilize materials used in medicine. This differentiation of application is directly attributable to the lower penetration of electrons and the high dose required by most chemical processes.
Accelerators in current use for processing materials operate in a direct current mode. They consist of two rain classifications designated "electron curtain" machines where the energy is restricted to less than 500 keV and "high voltage" machines where the maximum energy is 5 MeV.
Recently, industrial linear accelerators have been developed which are able to accelerate electrons to 10 MeV with power levels up to 20 kW. They offer the prospect of allowing electron accelerators to enter the lucrative medical sterilization market. A feature of the higher energy is the ability to convert the electron energy to photons with an efficiency which is more than twice that possible with 5 MeV electrons. This property of the electron nuclear interactions is further enhanced by kinematic considerations which demand that the photon beam be projected more in the forward direction. This means that for a given beam power the photon flux on-axis is seven times more intense at 10 MeV than at 5 MeV.
All dc accelerators stand off the high voltage across an insulated accelerating tube which contains the accelerating electrodes. Electrons entering the tube are accelerated to the final energy determined by the terminal voltage. The weakness of this system is that under intense radiation, electric charges will be created on the insulating tube and breakdown can occur. This breakdown will also occur under the electrical stress of the field itself. This is a direct consequence of the fundamental principle that the final electron energy, as defined in electron volts, is set by the actual voltage which the insulator must withstand. In practice, for industrial accelerators the energy limit imposed by this limitation is 5 MeV. In pushing these limits, manufacturers are tempted to compromise reliability.
The linear accelerator (linac) does not suffer from this limitation. It consists of a copper tube with a series of specifically shaped discs or cavities along its length. The oscillating electric field is contained within this copper tube, which is held at ground potential. Depending on the frequency of oscillation and the gradient, the actual potential difference between any two points in the system does not exceed 500 keV. An insulator is not required to sustain the high electric fields associated with this voltage. Existing industrial linacs work under a high level of stress which is undesirable to an industrial machine. This is a direct consequence of their historical pedigree rooted in particle physics research where emphasis is on high energy, high peak power, high field gradient and high klystron voltage with lesser consideration to high average power. The present invention addresses all of these limitations.
The present invention provides a new type of industrial linear accelerator that is conservatively inside the performance limits of accelerator technology. Energy gradients of research and medical linacs are typically 10 MeV/m. The gradient of the present invention is 3 MeV/m. Average power gradients have been tested in operational electron linacs of 100 kW/m. The present invention provide gradients of 15 kW/m. Beam currents during the pulse are of the order of 1A in existing pulsed linacs while the present invention produces a beam current of about 100 mA during the pulse.
These conservative ratings are made possible by using an L-band single accelerator structure with a Wehnelt controlled electron gun, a graded-.beta. capture section directly coupled to a .beta.=1 section and by driving the assembly with a low-peak power, modulated-anode klystron operated in a long pulsed mode. The long pulse has several advantages including the requirement for very modest peak power (2.5 MW), consequent low voltages on the klystron (&lt;100 kV) and a modulated anode which provides the pulse structure without having to transfer the power as in a conventional line modulator. The modest beam current means that beam-cavity interactions, which commonly consume power by exciting beam break up (bbu) modes, are rendered impotent. These basic physics principles have been embodied into an engineered prototype which has operated at 10 MeV and 50 kW with an availability of over 97% for over 1500 hours of full power operation.
A very important aspect of the long pulse concept is the ability to use the length of pulse as a variable and hence vary the average power of the beam without changing the physics of the process. The field gradient, the peak power and the current all remain the same. To vary the power of the machine at a constant energy, only the pulse length need be adjusted.
The novel feature associated with the long pulse is the ability to control the energy of the accelerated electrons during the pulse. The energy gained by the electrons traversing the structure is the line integral of the electric field. The amplitude of the electric field is controlled using a magnetic field probe to extract some of the power of the cavity, using a crystal detector to measure the amplitude and, after comparing with a voltage setpoint, sending a signal to the rf drive of the klystron to adjust the klystron output. The setpoint thus becomes the accelerator energy setpoint that can be directly linked to an international standard. A major advantage of this method of energy control is the elimination of the need of a magnetic bend to determine the energy and to assure that the possibility of unwanted excursions is eliminated.
Existing industrial rf linear accelerators operate with short pulses whereby rf energy is transmitted to the accelerator in an open loop mode. In this mode, changes in beam current result in a change in the rf field level in the accelerator and hence in a change in energy. This is particularly true of accelerators that dominate the existing industrial rf linac market. In these accelerators, the power and energy are closely tied together and, as the power is increased, the energy must drop. This is a problem for many applications where a variation in the flow of product and, hence, the tram power is necessary but where the energy must remain fixed within tight limits.
Tight energy tolerances can be achieved with expensive power supplies requiring very high stability. These systems use a time avenge of many pulses to determine a setpoint on the power supply for the energy. They are susceptible to changes in the pulse repetition rate. It is not possible to change the energy during the period of a single pulse with existing technology in the industrial linac field. Alternatively, the beam may be deflected by a calibrated amount In a magnetic field. This provides good energy selection following acceleration of the beam. However, existing systems do not allow the energy to be tightly controlled against the voltage droop that inevitably occurs during a pulse nor do they allow an independent control of the energy and power of the accelerator.
The present invention overcomes these difficulties by operating the accelerator in a long pulse mode with a fast, active feedback loop that can control the rf field during the accelerator pulse. The long pulse length, a pulse greater than 50 .mu.s, can be achieved with a modulated anode klystron. This provides sufficient time to permit regulation of the drive power to the klystron and hence control the beam energy at the energy setpoint. The beam current, and hence the beam power, is controlled by a separate control loop independently of the energy.
The wide range of applications to which electron accelerators have been subjected has led to unique machines designed for specific applications. Each accelerator has its own set of replacement components. The purchase cost of an accelerator and its replacement parts is high because of the non-recurring engineering cost associated with each part and the cost of inventory parts held by a supplier is high.
By way of background, a linear accelerator structure is composed of a series of cavities in which microwave power is used to establish electromagnetic fields. The cavities are designed to concentrate the electric fields in a beam aperture region of the cavities to accelerate charged particles. The accelerating energy gradient in the cavities is typically 10 MeV/m. The device has poor reliability for industrial use beyond an energy gradient of 10 MeV/m because electrical breakdown in the cavities disrupts beam acceleration.
The parameters that determine the output beam energy are length of the accelerator structure and the electric field gradient. Beams of high-energy are obtained with several accelerator structures in series. The drawback of having several accelerator structures in series is the need for additional control systems. The phase of the microwave fields in each accelerator structure must be controlled to ensure that particles are maintained in synchronism with the accelerating fields throughout the accelerator. The microwave transmission characteristics of each accelerator structure depend on the dimensions and temperature of the device. These must also be controlled precisely during fabrication and operation to obtain the desired output beam energy. The relative microwave power level in the different accelerator structures must be controlled. The control system is further complicated because of the coupling between the control parameters of the machine: phase, microwave transmission, accelerating field amplitude and accelerated beam current. These contribute to the uniqueness of each linear accelerator and, consequently, to the high purchase cost of an accelerator and its replacement parts.
The present invention seeks to simplify the high-energy linear accelerator by adopting a modular approach to address several applications with the same basic components. This allows the use of a single accelerator structure to achieve beams of high energy and eliminates the need for controlling the phase and microwave transmission characteristics of a multi-structure linear accelerator.
In accordance with this aspect of the present invention, the accelerator structure is composed of three building sections: a beam capture section module, a coupler section module and an acceleration section module. The length and number of these modules, joined together to form a monolith accelerator structure, are chosen to meet the desired beam energy and power for a particular application. A family of high-energy accelerators which can address different applications, using the same building components, can then be made available.
The capture section is designed to accelerate and form beam bunches synchronized with the microwave accelerating fields. The coupler section is a device used to transmit the microwave power into the accelerator structure. The acceleration section is composed of a series of identical cavities in which microwave power is used to accelerate the beam. Accelerator sections are joined together with flanges designed to establish good electrical contact for the flow of microwave current and to provide an ultra-high vacuum seal. This is achieved by compressing a copper gasket between two pairs of stainless steel knife edges. The inner pair of knife edges are used for the electrical contact and the outer pair of knife edges are used for the ultra-high vacuum seal.
The cross-sectional area of the electron beam leaving a high power irradiator must be large to ensure good spot overlap during scanning. This is accomplished with the L-band accelerating system. Also, a uniform dose distribution is required at the product to be irradiated.
The dose distribution is governed by software generated waveforms loaded into an arbitrary function generator. Output from the signal generator controls a bipolar power supply which drives the scanning electromagnet.
The electric field strength within a long-pulse linac must be regulated to within a few percent despite changes in beam loading and significant changes in the rf system gain. This regulation must be maintained on a microsecond time scale during the pulsed application of rf power. Regulation is also maintained from pulse to pulse. Good regulation is required to achieve predictable and reproducible irradiator performance. It is also beneficial in that overall electrical efficiency is improved by maintaining a preset beam energy and avoiding beam spill that results from energy-optics mismatch.
Heretofore, electric field regulation was achieved by using short pulses and time-averaged control. Use of short pulses prevents the rapid drop of rf gain from having an appreciable effect within a pulse. Pulse-to-pulse regulation is not done, rather the field strength is averaged over many pulses and controlled to a setpoint. As indicated, this method does not provide any intra-pulse regulation. When longer pulses are present, adaptive waveform-shaping has been used in which the error observed during a pulse is used to correct the input drive signal for the following pulse. This method requires complex digital signal processing circuits.
The present invention proposes a controller which consists of broadband yet simple proportional-integral analog control electronics and a single analog to digital converter (ADC) configured as a zero-droop sample and hold. An integration term is applied after a predetermined delay from the start of each pulse. After another short time-delay, the control signal is sampled and stored in the ADC. At the end of the pulse, the integration term is zeroed. At the start of the next pulse, the control signal is set to the value stored in the ADC and the proportional control term is engaged. The cycle repeats for each pulse. The method provides both fixed intra-pulse regulation and pulse-to-pulse regulation with simple electronics. Storing the control signal for use on the subsequent pulse and the staged deployment of the controller terms, effectively removes the dead-time between pulses, thus attaining the performance of a continuous system with a pulsed system.
The power for a pulsed electrical load is often derived from the electrical energy stored in a capacitor bank. The high discharge pulse current generally causes the voltage on the capacitor to droop significantly during the pulse, thereby changing the operation of the driven load during this time. A klystron is an example of such a driven load and a klystron with a modulating anode is often driven by a circuit which includes a switch, a pull-down resistor and the capacitor bank to store the charge for the current pulse through the klystron. When the switch closes, the klystron conducts current and can be used to amplify rf power. The declining voltage during the pulse affects both the cathode potential and the modulated anode potential in such a manner that the accelerating potential, i.e. the difference between the two, changes during the pulse. This circuit is not adequate if a controlled, predetermined change in the accelerating potential is desired.
It has been proposed to employ a programmable variable-voltage power supply to achieve a controlled accelerating potential. The power supply would be commanded to change its output voltage in a predetermined manner during the pulse. This system has proven to be costly and susceptible to reliability problems due to its complexity and number of active components.
The present invention proposes the provision of a switch tube triggered by a low power switch in order to divert a part of the current that flows through the resistor during the pulse through a grid-leak resistor in the switch tube circuit and from there through a diode to a small capacitor connected to ground. With the current during the pulse flowing through the capacitor, the magnitude of the voltage on the capacitor will decrease, drawing the modulated anode voltage with it. By the proper choice of grid-leak resistor, capacitor and the output impedance of the bias supply, the rate of voltage decrease during the pulse can be set to a predetermined value. Although this implementation involves the use of a switch tube, it will be understood that the same principle can be used with transistors as switching elements.
Control of the temperature of an accelerator gun cathode is required in order to maintain the cathode electron emission at it sufficiently high value and to prevent over-heating from damaging the cathode or shortening its life. Accelerator electron gun cathodes ate operated at elevated temperatures (&gt;1000.degree. C.) with heating provided by electrical current in a filament heater circuit. Depending upon the cathode type, the electron emission for a given electric potential distribution increases with increasing temperature. This emission characteristic is non-linear, approaching saturation at and above the operating temperature. Operation at excessive temperatures shortens the life of the cathode and increases the risk of gun arcing due to deposition of cathode material on insulating surfaces.
Radio-frequency linear accelerators accept injected electrons for forward acceleration and reject a fraction of the injected electrons. For accelerators not having a beach "buncher", the rejected electrons may be returned to the gun with significantly greater energy than they had on injection. This backwards-accelerated beam represents a small power loss to the accelerator and a significant power source to the electron gun. For an axi-symmetric geometry, a fraction of the backwards-accelerated electrons will impact on the gun cathode, deposit their energy and increase its temperature. Depending on the injection voltage and injection optics, this rejected beam may become a significant fraction of the power supplied to the cathode heater, altering the operating conditions.
In addition to the backwards accelerated electron beam, the accelerator will also accelerate ions generated from the background gas present in the accelerator. While the accelerator is not optimized for ion acceleration, some ion bombardment will occur. The gas present in the electron gun is ionized by the injected electron beam and the backwards accelerated beam produces a "column" of ions in front of the cathode. These ions will be accelerated by the cathode potential to impact the cathode and other surfaces at negative potential.
For most applications developed to date, the average backwards accelerated beam power is a small faction of the cathode heater power due to the low duty cycle (low average beam power) of the accelerator. Where mitigating measures are required (electron tubes), hollow cathode constructions have been employed or proposed to reduce the portion of the reverse beam impinging on the cathode. In addition, occluding optics may be employed to reduce the portion of the backwards accelerated beam that impacts the cathode. Moreover, it is possible to reduce the energy of the electrons returning to the cathode by operating the cathode at a greater injection voltage, requiting the electrons to "climb the coulomb barrier" before reaching the cathode.
As the average power of the accelerator is increased, the fraction of the cathode heater power that the power deposited by the backward accelerated beams represents grows to become significant. Adjustment of the injection optics by either mechanical or electromagnetic means reduces the back-heating fraction, but does not eliminate the phenomenon. At some finite average power, the back-heating effects prove limiting to further increases in average beam power without deleterious consequences.
The present invention estimates circuit resistance based on measurements of the gun cathode filament circuit voltage and current. A control loop is used to maintain the resistance at a setpoint value by adjusting the filament power supply current setpoint. This control loop may be implemented either in hardware or as a software control program of the accelerator. The filament circuit resistance serves to stabilize the cathode temperature and hence the electron gun performance under the influence of backward accelerated beam and/or ion bombardment. This resistance is used as an imperfect monitor of the cathode temperature.
Fast shutdown systems are required for linear accelerators to protect high power subsystems from damage. In particular, the shutdown systems are required to discharge the electrical energy stored in the rf power system in the event of anomalous conditions, to extinguish arcs in the rf power delivery system, preventing damage to the waveguide and components, to extinguish arcs in the linear accelerator, minimizing damage to the interior of the accelerator and protecting the rf power system from reflected power, to prevent anomalous rf drive conditions from damaging expensive components, to prevent deposition of excessive accelerated beam current on sensitive elements of the accelerator beam delivery system, and to disable accelerated beam current in the event of a failure of the beam dispersal subsystem.
The topology of a modem high-power accelerator has the major components distributed as appropriate to the requirements of the facility. In such a facility, the components that contribute to the decision that a fault condition exists may be separated from each other as well as from the logical point of action for the decision. The speed of decision and maximum delay to the protective action required are different depending on the characteristics of the fault condition and the tolerance of the affected components for the resulting stress. In many cases, the speed of detection and action exceeds the capabilities of the process control system by several orders of magnitude: a few microseconds as opposed to tens or hundreds of milliseconds. Hence, fast hard-wired protection systems are required.
Conventional protection practice depends, in pan, on the design of the accelerator and the limitations imposed by the component manufacturer. For example, until recently, most control systems have been arranged with each signal carried by individual wires to the control room for monitoring and alarm functions. Modern distributed control system designs permit reducing the number of signal cables that enter the control room, with most data being acquired remotely and telemetered via multiplexer digital communication from clustered points. An alternative practice is to provide a high speed detection function at the point of measurement, relay the decision to the control room where it may be logically conditioned and relay the instructions to the protective action point.
The multiple cables required for the conventional schemes carry cost penalties for the cable and installation, have multiple length signalling delays, and are vulnerable to the electromagnetic interference unless high cost optical-fibre systems are used. For specific types of faults, the associated electrical disturbance may be sufficient to defeat the communication function and to prevent protection. The system may also be vulnerable to spurious trips arising from external sources of electromagnetic interference.
These difficulties are overcome by the present invention by the provision of a single communication cable configured as a fail-safe current loop and used for high speed signalling of many protection decisions to one or more activation devices. The optically-isolated communication in the fail-safe sense is achieved with high speed by using a complementary logic drive to discharge the base capacitance of the primary optical isolator with a second optical isolator. The noise immunity for each decision is selected on the basis of the impact of the related fault condition permitting a unique false-alarm/missed-alarm tradeoff for each condition.
The high speed protection system of the present invention employs several key elements. It includes a current loop that is optically-isolated at each connection and chained through each decision device and action module. The current loop is enabled by the supervisory control system to permit testing and logical control. The current loop is arranged to be fail-safe in that a loss of continuity in the loop cable causes the action device to operate and the head-end control to latch the loop in an open state until it is reset. Decision modules employ the full sensor bandwidth available for detection and provide a selectable sustain criterion for the decision as well as limited provision for logical conditioning based on parameters monitored in other modules. A high quality digital communication cable is used for the current loop with the shield connections arranged for high noise immunity. Fault detection circuits are conditioned on the current loop being closed to ensure that, within the signalling delay, only the first fault to be detected is latched for diagnostic purposes. Each signal used for a protection function is separately measured by the supervisory process controller to validate the signal.